Generally speaking, Raman spectroscopy can provide molecular information via inelastic light scattering without physical contact. Coupled with microscopic imaging, Raman spectroscopy is a powerful technique for compositional analysis via inelastic light scattering. Raman microspectroscopy can be utilized for material analysis in many different ways that can include, for example, stress and temperature measurement in silicon and compositional analysis of polymer microparticles. Raman spectroscopy is particularly useful in obtaining microscopic information over a sizable area without physical contact, thereby enabling compositional analysis with little or no sample preparation. However, conventional laser scanning confocal Raman microscopy requires long data acquisition time due to its sequential operation, as well as the additional latency from the readout time of low-noise charge coupled device (CCD) detectors. As a result, conventional point-scan Raman mapping speed is about one to a few points per second, or a few Hertz.
Three methodologies are commonly employed in contemporary Raman microspectroscopy, namely, point-scan, line-scan, and global illumination. The point-scan operation involves the collection of Raman spectra in a point-by-point fashion. Since Raman scattering is a relatively weak phenomenon, the laser spot dwelling time at each point is typically on the order of milliseconds to seconds. In addition, the data needs to be read-out after each point acquisition, which may another few hundred milliseconds for a standard charge-coupled device (CCD) detector. As a result, conventional point-scan Raman mapping is a time-consuming process and, for example, can take as long as a few hours to map a 50×50 μm2 region.
To improve efficiency, parallel acquisition has been implemented based on time-sharing or power-sharing schemes. In the former, the laser is rapidly scanned over multiple points of interest during the time of a single CCD recording frame. In the latter, the laser is shaped into an elongated line and the entire line is imaged by a single CCD frame. Thus, both schemes can substantially reduce multiple read-out times. A key difference, however, lies in the temporal power fluctuation within a CCD frame. For time-sharing, the total laser power is focused on one spot at any given time but for power-sharing the laser power is distributed on all spots. Therefore, the frame-averaged power is identical to the instantaneous power for power-sharing but not for time-sharing. Another significant difference is there is no scanning within each frame for power-sharing. Recently, the time-sharing approach has been demonstrated to provide flexibility for imaging multiple points not aligned on a line, which is particularly advantageous for sparse samples such as bacteria or environmental particles.
To achieve a speed beyond 100 Hz, rapid point scanning may be employed in state-of-the-art commercial Raman systems with the use of electron-multiplied CCD (EMCCD) detectors. Although EMCCD can effectively overcome the read noise due to the short integration time per point measurement (˜msec), substantial noise is generated during the multiplication process. In addition, the throughput advantage can be severely undermined for common specimens, where the Raman scattering cross section is typically small, and thus requires much longer integration time to achieve a decent signal-to-noise ratio. In contrast, parallel acquisition using, for example, a line-shaped laser pattern can achieve similar throughput without the need for an EMCCD. Raman photons originating from the entire line, equivalent to many spots, are imaged to the entrance slit of a spectrograph, dispersed, and captured by a single CCD frame. Two-dimensional mapping is achieved by scanning the laser line in the transverse direction. Although efficient, the line-scan approach suffers from a major limitation: parallelism is only possible for points lying on a line, which severely undermines its throughput advantage when sparse sampling is desired.